Stack-and-stitch: ordinary optics vs telecentric

rjlittlefield · Post by **rjlittlefield** » Sat Jan 06, 2007 4:50 pm

Steve,

Um, not to put too a fine point on it, but those pictures you just posted (here)would make a lousy advertisement for whatever software made them.

What you've shown looks like a classic case of what happens without proper scaling and image alignment. Any software worth its salt should handle the situation with no problems.

To illustrate...

I took my favorite Autographa moth and set up a low mag, high DOF situation. 10 mm across the field, nominal f/16, 6 frames, focus step 0.020 inch. There's an easily visible scale change from frame to frame. (CombineZ5 reports 0.8%, which agrees with theory at 0.020 inch step and 2.5" from subject to entrance pupil.)

Here are the first and last frames, animated (total scale change 4%).

Here is a side-by-side comparison to show how much scaling matters. Left side, without scaling; right side, with scaling.

This happens to be with Helicon Focus, version 4.03, all autoadjustment turned off on the left, default parameters on the right. But there's nothing special about HF here -- CombineZ5 and Panorama Tools do just as well.

Based on what I've seen and heard so far, I'd say that something in your software setup is disabling scaling.

But I have no idea what that would be.

--Rik
(Edit to add link to earlier photos, appearing on previous page.)

svalley · Post by **svalley** » Sat Jan 06, 2007 5:34 pm

Rik, yep a classic case, of course, you have to remember that it had not occurred to me to look for a scaling routine in my software until we started talking. The ImagePro software has all sorts of image analysis and manipulation options that do not apply to imaging through a microscope, so I ignored most of them. When I started rummaging around there they were, all kinds of routines I could apply to my stack before compositing. They were in a completely unrelated section of the software so I did not realize they could be used on my stacks. I will have to try them out. This does show poor menu organization on the part of the people who designed the program. I hope the new versions will be better.

I still think I will prefer stitched tile composites for many of my images and I know my boss will. It is pretty hard to beat the resolution of fine detail that you get on an insect when the image has 25000 pixels on a side!

Thank you for all your comments and descriptions, etc. I am always trying to improve my technique and these discussions help a lot.

rjlittlefield · Post by **rjlittlefield** » Sat Jan 06, 2007 6:45 pm

svalley wrote:Rik, yep a classic case, of course, you have to remember that it had not occurred to me to look for a scaling routine in my software until we started talking.

Sorry if I sounded a little overly critical there. The images that you have put out are gorgeous. I'm even more impressed now, since it seems you've gotten them using software settings or procedures that maybe weren't the best.

This does show poor menu organization on the part of the people who designed the program.

I think you've hit the nail on the head about poor menu organization, or perhaps just menus that are poorly organized for this task. I'm a software R&D guy. In my world, one of the sayings that we use to console ourselves is "If we knew what we were doing, it wouldn't be research." We frequently find ourselves building stuff that has a multitude of little functions that can be combined in interesting and useful ways, and then the question becomes how to organize and package some of those ways so as to be more task oriented. More and more, lately, I find myself labeling the extremes of GUI's as being either "developer's interface" or "end-user interface". Helicon Focus has what I'd call an end-user interface -- simple, not many controls, and right out of the box it does the right thing if you just push the "go" button. I'm guessing that ImagePro has more of a developer's interface -- powerful and flexible, but requiring a hefty learning curve to get any specific thing done.

I still think I will prefer stitched tile composites for many of my images and I know my boss will. It is pretty hard to beat the resolution of fine detail that you get on an insect when the image has 25000 pixels on a side!

I agree. That's why I'm still wrestling with the question of how to get really correct geometry for stack and stitch. Telecentric optics are one promising route, but they have some big disadvantages too. Rotating about the lens entrance pupil would give wonderful results, but so far I haven't figured out how to do that without some nightmarish mechanical contraption. Shallow DOF and just don't scale may well turn out to be the best approach for the majority of subjects.

--Rik

svalley · Post by **svalley** » Sat Jan 06, 2007 7:23 pm

That's why I'm still wrestling with the question of how to get really correct geometry for stack and stitch. Telecentric optics are one promising route, but they have some big disadvantages too.

Rik, I think for guys like us at least half the fun is trying to come up with innovative solutions to the technical problems. I know I get an immense sense of satifaction when I end up with an image that portrays a difficult feature exactly the way I want. The test comes when I ask the other guys in my lab, "hey, what does this look like to you?" I also love it when I figure out some new technique in photoshop or the other software I use and I can say, "this is going to save me an hour a day." I guess the pursuit of perfection can be its own reward.

cactuspic · Post by **cactuspic** » Sun Jan 07, 2007 7:59 pm

Rik,

I sometimes attach my Canon to the back kof my 4x5 Sinar. Assuming a lens with a sufficient image circle, do you anticipate an problems for me if I stitch together two image stack from the same image circle. The proceedure I inttend to use is to run one stack, move the shift the back, and then run the second stack.

Irwin

rjlittlefield · Post by **rjlittlefield** » Mon Jan 08, 2007 12:28 am

Irwin, that's a good question.

My first response is "Yeah, that should work fine", based on the theory that both stacks will have the same entrance pupil position, or at least the same set of positions. But then I got to worrying about the details. To preclude smearing or echos, you'll have to enable auto-adjustment at least for scale. As soon as you do that, you lose any guarantee that both stacks will do exactly the same thing, though it should be close. So my second response is "Yeah, that'll probably work fine. At least it should be a lot better than moving the lens sideways between stacks."

--Rik

cactuspic · Post by **cactuspic** » Mon Jan 08, 2007 7:36 am

Rik,

I am hoping that by carefully indexing my start and end points, the stacks will behave. But why should they be better than my children

I will let you know.

Irwin